Transformer models have achieved outstanding results on a variety of language tasks, such as text classification, ma- chine translation, and question answering. This success in the field of Natural Language Processing (NLP) has sparked interest in the computer vision community to apply these models to vision and multi-modal learning tasks. However, visual data has a unique structure, requiring the need to rethink network designs and training methods. As a result, Transformer models and their variations have been suc- cessfully used for image recognition, object detection, seg- mentation, image super-resolution, video understanding, image generation, text-image synthesis, and visual question answering, among other applications.

We present BoTNet, a conceptually simple yet powerful backbone architecture that incorporates self-attention for multiple computer vision tasks including image classification, object detection and instance segmentation. By just replacing the spatial convolutions with global self-attention in the final three bottleneck blocks of a ResNet and no other changes, our approach improves upon the baselines significantly on instance segmentation and object detection while also reducing the parameters, with minimal overhead in latency. Through the design of BoTNet, we also point out how ResNet bottleneck blocks with self-attention can be viewed as Transformer blocks. Without any bells and whistles, BoTNet achieves 44.4% Mask AP and 49.7% Box AP on the COCO Instance Segmentation benchmark using the Mask R-CNN framework; surpassing the previous best published single model and single scale results of ResNeSt [67] evaluated on the COCO validation set. Finally, we present a simple adaptation of the BoTNet design for image classification, resulting in models that achieve a strong performance of 84.7% top-1 accuracy on the ImageNet benchmark while being up to 1.64x faster in "compute"1time than the popular EfficientNet models on TPU-v3 hardware. We hope our simple and effective approach will serve as a strong baseline for future research in self-attention models for vision.2

Self-attention has the promise of improving computer vision systems due to parameter-independent scaling of receptive fields and content-dependent interactions, in contrast to parameter-dependent scaling and content-independent interactions of convolutions. Self-attention models have recently been shown to have encouraging improvements on accuracy-parameter trade-offs compared to baseline convolutional models such as ResNet-50. In this work, we develop self-attention models that can outperform not just the canonical baseline models, but even the high-performing convolutional models. We propose two extensions to self-attention that, in conjunction with a more efficient implementation of self-attention, improve the speed, memory usage, and accuracy of these models. We leverage these improvements to develop a new self-attention model family, HaloNets, which reach state-of-the-art accuracies on the parameter-limited setting of the ImageNet classification benchmark. In preliminary transfer learning experiments, we find that HaloNet models outperform much larger models and have better inference performance. On harder tasks such as object detection and instance segmentation, our simple local self-attention and convolutional hybrids show improvements over very strong baselines. These results mark another step in demonstrating the efficacy of self-attention models on settings traditionally dominated by convolutions.1

Self-attention has recently been adopted for a wide range of sequence modeling problems. Despite its effectiveness, self-attention suffers from quadratic computation and memory requirements with respect to sequence length. Successful approaches to reduce this complexity focused on attending to local sliding windows or a small set of locations independent of content. Our work proposes to learn dynamic sparse attention patterns that avoid allocating computation and memory to attend to content unrelated to the query of interest. This work builds upon two lines of research: It combines the modeling flexibility of prior work on content-based sparse attention with the efficiency gains from approaches based on local, temporal sparse attention. Our model, the Routing Transformer, endows self-attention with a sparse routing module based on online k-means while reducing the overall complexity of attention to O( n 1.5 d) from O( n 2 d) for sequence length n and hidden dimension d. We show that our model outperforms comparable sparse attention models on language modeling on Wikitext-103 (15.8 vs 18.3 perplexity), as well as on image generation on ImageNet-64 (3.43 vs 3.44 bits/dim) while using fewer self-attention layers. Additionally, we set a new state-of-the-art on the newly released PG-19 data-set, obtaining a test perplexity of 33.2 with a 22 layer Routing Transformer model trained on sequences of length 8192. We open-source the code for Routing Transformer in Tensorflow. 1

Advances in learning and representations have reinvigorated work that connects language to other modalities.A particularly exciting direction is Vision-and-Language Navigation (VLN), in which agents interpret natural language instructions and visual scenes to move through environments and reach goals.Despite recent progress, current research leaves unclear how much of a role language understanding plays in this task, especially because dominant evaluation metrics have focused on goal completion rather than the sequence of actions corresponding to the instructions.Here, we highlight shortcomings of current metrics for the Room-to-Room dataset (Anderson et al., 2018b) and propose a new metric, Coverage weighted by Length Score (CLS).We also show that the existing paths in the dataset are not ideal for evaluating instruction following because they are direct-to-goal shortest paths.We join existing short paths to form more challenging extended paths to create a new data set, Room-for-Room (R4R).Using R4R and CLS, we show that agents that receive rewards for instruction fidelity outperform agents that focus on goal completion.

Convolutional networks have enjoyed much success in many computer vision applications. The convolution operation however has a significant weakness in that it only operates on a local neighbourhood, thus missing global information. Self-attention, on the other hand, has emerged as a recent advance to capture long range interactions, but has mostly been applied to sequence modeling and generative modeling tasks. In this paper, we propose to augment convolutional networks with self-attention by concatenating convolutional feature maps with a set of feature maps produced via a novel relative self-attention mechanism. In particular, we extend previous work on relative self-attention over sequences to images and discuss a memory efficient implementation. Unlike Squeeze-and-Excitation, which performs attention over the channels and ignores spatial information, our self-attention mechanism attends jointly to both features and spatial locations while preserving translation equivariance. We find that Attention Augmentation leads to consistent improvements in image classification on ImageNet and object detection on COCO across many different models and scales, including ResNets and a state-of-the art mobile constrained network, while keeping the number of parameters similar. In particular, our method achieves a 1.3% top-1 accuracy improvement on ImageNet classification over a ResNet50 baseline and outperforms other attention mechanisms for images such as Squeeze-and-Excitation. It also achieves an improvement of 1.4 AP in COCO Object Detection on top of a RetinaNet baseline.

Convolutions are a fundamental building block of modern computer vision systems. Recent approaches have argued for going beyond convolutions in order to capture long-range dependencies. These efforts focus on augmenting convolutional models with content-based interactions, such as self-attention and non-local means, to achieve gains on a number of vision tasks. The natural question that arises is whether attention can be a stand-alone primitive for vision models instead of serving as just an augmentation on top of convolutions. In developing and testing a pure self-attention vision model, we verify that self-attention can indeed be an effective stand-alone layer. A simple procedure of replacing all instances of spatial convolutions with a form of self-attention to ResNet-50 produces a fully self-attentional model that outperforms the baseline on ImageNet classification with 12% fewer FLOPS and 29% fewer parameters. On COCO object detection, a fully self-attention model matches the mAP of a baseline RetinaNet while having 39% fewer FLOPS and 34% fewer parameters. Detailed ablation studies demonstrate that self-attention is especially impactful when used in later layers. These results establish that stand-alone self-attention is an important addition to the vision practitioner's toolbox.

Mia Xu Chen, Orhan Firat, Ankur Bapna, Melvin Johnson, Wolfgang Macherey, George Foster, Llion Jones, Mike Schuster, Noam Shazeer, Niki Parmar, Ashish Vaswani, Jakob Uszkoreit, Lukasz Kaiser, Zhifeng Chen, Yonghui Wu, Macduff Hughes. Proceedings of the 56th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2018.

Autoregressive sequence models based on deep neural networks, such as RNNs, Wavenet and Transformer are the state-of-the-art on many tasks. However, they lack parallelism and are thus slow for long sequences. RNNs lack parallelism both during training and decoding, while architectures like WaveNet and Transformer are much more parallel during training, but still lack parallelism during decoding. We present a method to extend sequence models using discrete latent variables that makes decoding much more parallel. The main idea behind this approach is to first autoencode the target sequence into a shorter discrete latent sequence, which is generated autoregressively, and finally decode the full sequence from this shorter latent sequence in a parallel manner. To this end, we introduce a new method for constructing discrete latent variables and compare it with previously introduced methods. Finally, we verify that our model works on the task of neural machine translation, where our models are an order of magnitude faster than comparable autoregressive models and, while lower in BLEU than purely autoregressive models, better than previously proposed non-autogregressive translation.

Image generation has been successfully cast as an autoregressive sequence generation or transformation problem. Recent work has shown that self-attention is an effective way of modeling textual sequences. In this work, we generalize a recently proposed model architecture based on self-attention, the Transformer, to a sequence modeling formulation of image generation with a tractable likelihood. By restricting the self-attention mechanism to attend to local neighborhoods we significantly increase the size of images the model can process in practice, despite maintaining significantly larger receptive fields per layer than typical convolutional neural networks. While conceptually simple, our generative models significantly outperform the current state of the art in image generation on ImageNet, improving the best published negative log-likelihood on ImageNet from 3.83 to 3.77. We also present results on image super-resolution with a large magnification ratio, applying an encoder-decoder configuration of our architecture. In a human evaluation study, we find that images generated by our super-resolution model fool human observers three times more often than the previous state of the art.

Batch-splitting (data-parallelism) is the dominant distributed Deep Neural Network (DNN) training strategy, due to its universal applicability and its amenability to Single-Program-Multiple-Data (SPMD) programming. However, batch-splitting suffers from problems including the inability to train very large models (due to memory constraints), high latency, and inefficiency at small batch sizes. All of these can be solved by more general distribution strategies (model-parallelism). Unfortunately, efficient model-parallel algorithms tend to be complicated to discover, describe, and to implement, particularly on large clusters. We introduce Mesh-TensorFlow, a language for specifying a general class of distributed tensor computations. Where data-parallelism can be viewed as splitting tensors and operations along the "batch" dimension, in Mesh-TensorFlow, the user can specify any tensor-dimensions to be split across any dimensions of a multi-dimensional mesh of processors. A Mesh-TensorFlow graph compiles into a SPMD program consisting of parallel operations coupled with collective communication primitives such as Allreduce. We use Mesh-TensorFlow to implement an efficient data-parallel, model-parallel version of the Transformer sequence-to-sequence model. Using TPU meshes of up to 512 cores, we train Transformer models with up to 5 billion parameters, surpassing SOTA results on WMT'14 English-to-French translation task and the one-billion-word Language modeling benchmark. Mesh-Tensorflow is available at https://github.com/tensorflow/mesh

The dominant sequence transduction models are based on complex recurrent orconvolutional neural networks in an encoder and decoder configuration. The best performing such models also connect the encoder and decoder through an attentionm echanisms. We propose a novel, simple network architecture based solely onan attention mechanism, dispensing with recurrence and convolutions entirely.Experiments on two machine translation tasks show these models to be superiorin quality while being more parallelizable and requiring significantly less timeto train. Our single model with 165 million parameters, achieves 27.5 BLEU onEnglish-to-German translation, improving over the existing best ensemble result by over 1 BLEU. On English-to-French translation, we outperform the previoussingle state-of-the-art with model by 0.7 BLEU, achieving a BLEU score of 41.1.

Ashish Vaswani, Yonatan Bisk, Kenji Sagae, Ryan Musa. Proceedings of the 2016 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies. 2016.

Boliang Zhang, Xiaoman Pan, Tianlu Wang, Ashish Vaswani, Heng Ji, Kevin Knight, Daniel Marcu. Proceedings of the 2016 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies. 2016.

Barret Zoph, Ashish Vaswani, Jonathan May, Kevin Knight. Proceedings of the 2016 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies. 2016.

Transition-based approaches based on local classification are attractive for dependency parsing due to their simplicity and speed, despite producing results slightly below the state-of-the-art. In this paper, we propose a new approach for approximate structured inference for transition-based parsing that produces scores suitable for global scoring using local models. This is accomplished with the introduction of error states in local training, which add information about incorrect derivation paths typically left out completely in locally-trained models. Using neural networks for our local classifiers, our approach achieves 93.61% accuracy for transition-based dependency parsing in English.

Qing Dou, Ashish Vaswani, Kevin Knight, Chris Dyer. Proceedings of the 53rd Annual Meeting of the Association for Computational Linguistics and the 7th International Joint Conference on Natural Language Processing (Volume 1: Long Papers). 2015.

We present Model Invertibility Regularization (MIR), a method that jointly trains two directional sequence alignment models, one in each direction, and takes into account the invertibility of the alignment task.By coupling the two models through their parameters (as opposed to through their inferences, as in Liang et al.'s Alignment by Agreement (ABA), and Ganchev et al.'s Posterior Regularization (PostCAT)), our method seamlessly extends to all IBMstyle word alignment models as well as to alignment without parallel data.Our proposed algorithm is mathematically sound and inherits convergence guarantees from EM.We evaluate MIR on two tasks: (1) On word alignment, applying MIR on fertility based models we attain higher F-scores than ABA and PostCAT.(2) On Japanese-to-English backtransliteration without parallel data, applied to the decipherment model of Ravi and Knight, MIR learns sparser models that close the gap in whole-name error rate by 33% relative to a model trained on parallel data, and further, beats a previous approach by Mylonakis et al.

Inspired by previous work, where decipherment is used to improve machine translation, we propose a new idea to combine word alignment and decipherment into a single learning process.We use EM to estimate the model parameters, not only to maximize the probability of parallel corpus, but also the monolingual corpus.We apply our approach to improve Malagasy-English machine translation, where only a small amount of parallel data is available.In our experiments, we observe gains of 0.9 to 2.1 Bleu over a strong baseline.

Many statistical models for natural language processing exist, including context-based neural networks that (1) model the previously seen context as a latent feature vector, (2) integrate successive words into the context using some learned representation (embedding), and (3) compute output probabilities for incoming words given the context.On the other hand, brain imaging studies have suggested that during reading, the brain (a) continuously builds a context from the successive words and every time it encounters a word it (b) fetches its properties from memory and (c) integrates it with the previous context with a degree of effort that is inversely proportional to how probable the word is.This hints to a parallelism between the neural networks and the brain in modeling context (1 and a), representing the incoming words (2 and b) and integrating it (3 and c).We explore this parallelism to better understand the brain processes and the neural networks representations.We study the alignment between the latent vectors used by neural networks and brain activity observed via Magnetoencephalography (MEG) when subjects read a story.For that purpose we apply the neural network to the same text the subjects are reading, and explore the ability of these three vector representations to predict the observed word-by-word brain activity.Our novel results show that: before a new word i is read, brain activity is well predicted by the neural network latent representation of context and the predictability decreases as the brain integrates the word and changes its own representation of context.Secondly, the neural network embedding of word i can predict the MEG activity when word i is presented to the subject, revealing that it is correlated with the brain's own representation of word i.Moreover, we obtain that the activity is predicted in different regions of the brain with varying delay.The delay is consistent with the placement of each region on the processing pathway that starts in the visual cortex and moves to higher level regions.Finally, we show that the output probability computed by the neural networks agrees with the brain's own assessment of the probability of word i, as it can be used to predict the brain activity after the word i's properties have been fetched from memory and the brain is in the process of integrating it into the context.

We explore the application of neural language models to machine translation.We develop a new model that combines the neural probabilistic language model of Bengio et al., rectified linear units, and noise-contrastive estimation, and we incorporate it into a machine translation system both by reranking k-best lists and by direct integration into the decoder.Our large-scale, large-vocabulary experiments across four language pairs show that our neural language model improves translation quality by up to 1.1 Bleu.

We describe a spoken dialogue system which can engage in Call For Fire (CFF) radio dialogues to help train soldiers in proper procedures for requesting artillery fire missions. We describe the domain, an information-state dialogue manager with a novel system of interactive information components, and provide evaluation results. Index Terms: spoken dialogue systems. 1.